1
A Business & Clinical Case for Continuous
Surveillance
Session 209, February 14, 2019
Leah Baron, MD, Anesthesiologist
John Zaleski, PhD, CAP, CPHIMS, Chief Analytics Officer, Bernoulli Health
2
Conflict of Interest
Leah Baron, MD
Principal investigator and clinical lead of study.
3
John Zaleski, PhD
Employed by Bernoulli Health developer and manufacturer of
software & hardware for data capture employed in the study.
Conflict of Interest
4
The Challenge
Capnography for Continuous Monitoring
The Pilot
Alarm Fatigue
IRB Study Findings
Summary
Agenda
5
Identify the perioperative care challenges from technical, clinical, and
workflow perspectives associated with monitoring and identifying patients
either at-risk for or experiencing respiratory depression
Analyze the use of continuous monitoring to detect potential signs of
respiratory depression while differentiating non-actionable from actionable
alarm signals, and understand the impact on clinical alarm fatigue
Assess the creation and use of multivariate alarm signals using real-time
data to generate more clinically actionable notifications to clinical staff
Investigate the strategies, technologies and workflow processes required
for continuous patient monitoring and remote patient surveillance in a
general care floor setting
Identify opportunities and challenges associated with scaling continuous
patient monitoring and surveillance across the enterprise
Learning Objectives
6
Please complete online session evaluation
Leah Baron, MD
Anesthesiologist
609-670-6000 | www.linkedin.com/in/leah-baron-14b5b818
John Zaleski, PhD, CAP, CPHIMS
Chief Analytics Officer, Bernoulli Health
484-319-7345 | www.linkedin.com/in/johnzaleski
Questions
7
General Acknowledgement
What follows and the studies conducted could only have
been performed with the contributions of multiple individuals
and departments within the health system. The names are
many, and roles include nursing and nurses aids, telemetry
technicians, respiratory therapy, physicians, information
technology, biomedical/clinical engineering, & vendors
8
The Challenge
Improving patient safety for post-operative, at-risk patients
9
Improving post-operative patient safety
Caring for increasingly complex, co-morbid patients
Bariatrics, joint replacements, etc.
Co-morbidities: apnea, COPD, diabetes, heart failure,
etc.
Improving clinical oversight of post-operative quality of care
Reduce incidents of Opioid-Induced Respiratory Depression
(OIRD) on general care floor
Seeking ways to improve clinical oversight of at-risk patients
Continuous surveillance of key parameters
Notification paradigm to clinical stakeholders
The Challenge
10
Background
Pain management in the post-operative general care
environment
11
“Clinical deterioration on the general hospital wards is common ... all
too often results in patients progressing to cardiopulmonary arrest,
which carries significant morbidity and mortality
“Surgical patients may be prone to cardiopulmonary arrest due to their
underlying diseases (especially conditions such as obstructive sleep
apnea and cardiac disease)…
“This is a major concern for anesthesia professionals who help
determine if the patient is safe to go to an unmonitored floor or should
request a monitored bed which may be a limited resource…
Source: B.D. Winters: “Early Warning Systems: ‘Found Dead in Bed’ Should be a Never Event”
https://www.apsf.org/article/early-warning-systems-found-dead-in-bed-should-be-a-never-event/
Recent Literature
12
“…the frequency of vital sign acquisition on the wards may be
insufficient to allow detection of clinical deterioration…
“…Electronic health records (EHRs) do little to improve this situation
as their performance is dependent on the intermittently and
sometimes inaccurately collected data points…”
“Surveillance monitoring may be a better way to collect and act on
clinical data for a patient who is deteriorating on a general ward…
“…vital-sign data collection should be continuous, since use of
intermittently collected data may miss early signs of deterioration.”
Source: B.D. Winters: “Early Warning Systems: ‘Found Dead in Bed’ Should be a Never Event”
https://www.apsf.org/article/early-warning-systems-found-dead-in-bed-should-be-a-never-event/
Recent Literature
13
Up to 14% of patients using opioids suffer from respiratory
depression. ¹
50% of CODE BLUE events involve patients who recieve opioid
analgesia ²
Respiratory failure caused by acute hypercapnia can occur in a
matter of minutes. ³
Carbon Dioxide Narcosis
¹Maddox, R. R., & Williams, C. K. (2012). Clinical experience with capnography monitoring for PCA
patients. Anesthesia Patient Safety Foundation Newsletter, 1-7.
²Overdyk, F. J. (2010). Postoperative opioids remain a serious patient safety threat.
Anesthesiology, 113(1), 259-260.
³Kaynar, A. M., & Pinsky, M. R. (2012). Respiratory Failure. Retrieved from
http://emedicine.medscape.com/article/167981-overview
14
15
Source: APSF Newsletter, February 2018
APSF Newsletter, February 2018
16
17
Source:
”APSF Highlights 12
Perioperative Patient
Safety Priorities For 2018
https://www.apsf.org/articl
e/apsf-highlights-12-
perioperative-patient-
safety-priorities-for-2018/
18
19
20
Capnography for Continuous
Monitoring (Virtua Project)
Ventilation monitoring using capnography to detect
immediate changes in respiratory issues
21
Sedation Level
Use modified POSS
Monitor
q ½ hour x 2 on initiation of therapy
q 1 hour until stable as indicated by achievement of comfort
q 4 hours once stable
Respiration Rate
count for a full minute
assess rhythm and depth of chest excursion
assess for pauses in breathing pattern
Oxygenation: Pulse oximetry
Do not rely on pulse oximetry alone because pulse oximetry can suggest adequate
oxygenation in patients who are actively experiencing respiratory depression, especially
when supplemental oxygen is being used.
Ventilation: Capnography
Measures carbon dioxide in the exhaled breath
Patient Assessment for Patients
Receiving Narcotics
Source: “Preparing to Use Capnography:
Pre-Capnography Class Nursing Education
September, 2013”
22
Difference Between Capnography &
Pulse Oximetry
Source: “Preparing to
Use Capnography:
Pre-Capnography
Class Nursing
Education September,
2013”
23
Source: “Preparing to
Use Capnography: Pre-
Capnography Class
Nursing Education
September, 2013”
24
Initial Rollout…2013
Employ capnography to detect opioid-related respiratory
insufficiency in post-operative patients, using remote
telemetry monitoring
25
Description
2 hospital sites Memorial and Marlton
4NE, 4N
2 weeks- Mon- Fri (controlled environment)
Patient association in the PACU, via bar-coding
Transferred patients to floor
Remote monitoring to tele room for ~ 24 hours
Disassociation in Central Processing
Pilot Overview
26
Target population for capnography:
 
Significant:
COPD
Heart Failure
Cardiomyopathy
Emphysema

OSA presence
Concomitant use of CNS depressant meds (benzodiazepines)
Persistent hypoxemia
Patients requiring escalating doses of PCA, bolus or basal
Patients with known increased sensitivity to narcotics
Patients with history of chronic high doses of narcotics
Prolonged anesthesia times > 5 hours
Initial Pilot Rollout
27
STOP-BANG CRITERIA
28
Vital Signs Limit Thresholds
Monitoring Alarm Signal Limit Settings
Urgent Caution
ETCO2 High 50 45
ETCO2 Low 29 35
FiCO2 High 8 2
RR High 24 21
RR Low 6 8-10
No Breath
Sec 30
SpO2 High 100 100
SpO2 Low 91 95
PR High 110 100
PR Low 54 60
SAT Sec 25
IPI Low 3 5
29
Capnostream 20 Capnography Monitor
Capnogram: Wave form
Capnometer:
Numeric
measurement of
End-tidal CO
awRR:
Airway
Respiratory
Rate
Oxygen Saturation
Heart Rate
IPI-Integrated
Pulmonary Index:
a single number
that describes the
patients
respiratory status
Sampling
Line
Source: “Preparing to
Use Capnography:
Pre-Capnography
Class Nursing
Education September,
2013”
30
Tele-Tech Role:
Workflow
Decrease variation
Assessment and impact
31
Pop-Up Notifications: Bed-Board View
32
Urgent Alert
URGENT ALERT
with SOUND
33
CO2 Pump Paused
Pump
Paused
34
Cautionary Alert
CAUTION
ALERT
35
Alarm Print For Tele Workflow
36
Data Continuously Measured
0
10
20
30
40
50
60
70
80
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
CO2-EX (mmHg)
92
94
96
98
100
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
SPO2 (%)
Time (minutes)
37
Data Continuously Measured
0
5
10
15
20
25
30
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
RR (breath/min)
0
10
20
30
40
50
60
70
80
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
etCO2 (mmHg)
Time (minutes)
38
Significant Alarm Signals
0
5
10
15
20
25
30
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
RR (breath/min)
0
10
20
30
40
50
60
70
80
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
etCO2 (mmHg)
Time (minutes)
Signal threshold
breaches for
individual
measurements
became
overwhelming
39
Some Improvement: Sustained Signal
Delays
0
10
20
30
40
50
60
70
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
CO2-EX (mmHg)
CO2-EX 30 Sec
Running Ave
Time (minutes)
Raw data may have
aberrations leading to false
alarmsthese would
cause notifications to be
communicated when
thresholds are exceeded.
Running 30-sec average
establishes trend over 5
measurements (each
measurement 6 seconds
apart) so that when a
threshold is exceeded, is
occurring on basis of
multiple valuesmore
indicative of a real trend
etCO2 (mmHg)
etCO2 (mmHg)
30-s running ave
40
Communication of monitoring alarms directly to telemetry
technicians impractical
Overwhelming quantities of non-actionable notifications
Signals interleaved with artifact made identifying truly
actionable events next to impossible
Aside: No industry-standard alarm recommendations for
adoption at the time
Decision: explore the field of “combinatorial” machine alarms to
help reduce non-actionable alarm load
Tightly-controlled environment, with institutional oversight
With assistance from research nurses to add a layer of
safety to our patients
Pilot 2013 Findings
41
Alarm Fatigue
Non-actionable notifications that have no clinically-relevant
cause
42
False Alarm Rate
as high as 99%!
350 cardiac alarms
/ patient / day (ICU)
ECRI Institute. Top 10 Health Technology Hazards Report for
2014. https://www.ecri.org/press/Pages/2014-Top-10-Health-
Technology-Hazards-Report.aspx. Published November 4,
2014. 10 February, 2018.
Default machine-issued alarms are
not very skillful in identifying
adverse events.
43
MIMIC II Database (Beth Israel
Deaconess)
962 ICU patient sample;
SpO2 alarm rate threshold simulated
by applying range from 84%, 86%,
88%, 90%.
SpO2 alarm rates vary by patient,
time;
Affected by alarm settings;
Personalized settings may help
reduce alarm fatigue.
44
IRB Study…2015
Evaluate the use of smart rules in the environment on a
smaller cohort of patients & evaluate outcomes
45
Study Parameters
Conducted with IRB oversight
Limit to a specific hospital
general medical surgical unit
Obtain patient consent prior
to surgery
Identify patients particularly
at-risk for respiratory
depression
46
Study System Architecture
47
Capnostream Monitor
Wireless Serial Bridge
Sources: Supe et al. 2017; photos by author
Alarm Notifications via VoIP Phone
48
0
5
10
15
20
25
30
35
40
45
50
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
etCO2 (mmHg)
Time (minutes)
Non-consecutive etCO2 measurements which exceed threshold
threshold
Alarm Signals Based on Threshold
Breaches
49
Alarm Condition
ALR-DISC-SPO2 8527
ALR-LO-BAT 8301
ALR-NO-BREATH 7065
ALR-LO-IPI 5153
ALR-LO-CO2EX 3065
ALR-FL-DISC-CO2 3051
ALR-PR-NF 1925
ALR-HI-RR 2173
ALR-CO2-PUMP-OFF 1768
ALR-LO-PR 1637
ALR-OFF-SPO2 1318
ALR-LO-RR 1116
ALR-LO-SPO2 873
ALR-CO2-CHK-CAL 331
FL-BLOCK 308
ALR-STBY-CO2 264
ALR-CO2-CHK-FLW 52
CO2-MLFNC 45
SPO2-MLFNC 37
ALR-HI-PR 9
ALR-HI-CO2EX 3
ALR-HI-SPO2 0
ALR-STBY-SPO2 0
Total 47021
Limit Threshold Alarm Annunciations
Were Significant
50
0
5
10
15
20
25
30
35
40
45
50
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
etCO2 (mmHg)
Time (minutes)
Consecutive parameter measurements which exceed threshold
threshold
Alarm Signals Based on Sustained
Behavior (non-self-correcting)
51
Combinatorial Parameter Alarm
Annunciations Proved to be Most
Effective
52
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
160
170
0
1
5
3
0
4
5
6
0
7
5
9
0
1
0
5
1
2
0
1
3
5
1
5
0
1
6
5
1
8
0
1
9
5
2
1
0
2
2
5
2
4
0
2
5
5
2
7
0
PR (beats/minute)
time after arrival from operating room (minutes)
Caution
Urgent
Caution
Urgent
Example: Pulse versus time
53
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
0
1
5
3
0
4
5
6
0
7
5
9
0
1
0
5
1
2
0
1
3
5
1
5
0
1
6
5
1
8
0
1
9
5
2
1
0
2
2
5
2
4
0
2
5
5
2
7
0
SpO2 (%)
time after arrival from operating room (minutes)
Caution
Urgent
Example: Oxygen Saturation v. Time
54
Example: Respirations & etCO2 v. Time
55
0.0000
0.0500
0.1000
0.1500
0.2000
0.2500
0.3000
0.3500
0.4000
0.4500
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99100
SpO2 (%)
SPO2 Density Function
0.0000
0.0500
0.1000
0.1500
0.2000
0.2500
0 10 20 30 40 50 60 70 80 90 100 110 120 130
Pulse Rate (beats/minute)
PR Density Function
0.0000
0.0200
0.0400
0.0600
0.0800
0.1000
0.1200
0.1400
0.1600
0.1800
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
Respiratory Rate (breaths/minute)
RR Density Function
0.0000
0.0500
0.1000
0.1500
0.2000
0.2500
0.3000
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44
etCO2 (mmHg)
etCO2 Density Function
56
Use of Measurements to Create
Combination Alarm Signals: etCO2
Hypocapnia Alarm Limit
(CO2-EX < 15 mmHg)
Instantaneous & sustained
occurrences of
CO2-EX < 15 mmHg
57
Use of Measurements to Create
Combination Alarm Signals: Resp
Bradypnea Alarm Limit
(f
R
< 6 rpm)
Instantaneous & sustained
occurrences of
f
R
< 6 rpm
58
Use of Measurements to Create
Combination Alarm Signals: SpO2
Hypoxia Alarm Limit
(SpO2 < 90%)
Instantaneous & sustained
occurrences of
SpO2 < 90%
59
Single Combination Alarm
Possible instances of central or
obstructive sleep apnea: etCO2 = 0 &
f
R
= 0, with valid measurements
(i.e., no technical alarm thrown &
cannula properly in place on patient)
Reference:
J. Garah, O.E. Advi, I. Rosen, R. Shaoul,
“The value of Integrated Pulmonary Index (IPI)
monitoring during endoscopies in children” J
Clin Monit Comput (2015) 29:773-778
f
R
rpm
etCO2
mm Hg
13 individual alarms translate into 2 combined sustained alarms
60
Alarm Threshold Limits Considered
Alarm level changes or policies to reduce alarms must be
validated for clinical efficacy & safety
61
Correlations Among Machine-Reported
Alarm Signals
Low expired carbon dioxide (<15 mmHg)
Low respirations (< 6 per minute)
Low hemoglobin oxygen saturation (< 90%)
Low etCO2 Low
Respirations
Low SpO2
Low etCO2 1
Low Respirations 0.987 1
Low SpO2 0.892 0.885 1
62
Combination Alarms per Study
Combination of low respirations, oxygen saturation & end-tidal
carbon dioxide were chiefly correlated with patients who required
active interventions: truly apneic patients
63
3-Parameter Sensitivity & Specificity
(Small Patient Sample)
 


| Sensitivity: probability or fraction indicating that condition is present among those having the condition.
 


| Specificity: probability or fraction indicating a negative result in those not having the condition.
 

 
Positive Predictive Value (PPV): What is the probability that a patient with a positive result truly has condition?
Negative Predictive Value (NPV): What is the probability that a patient with a negative result truly does not have the condition?
 

 
Measure Value
Sensitivity 0.800
Specificity* 0.750
PPV 0.364
NPV* 0.955
Limited data
based on not
identifying
patients with
condition due to
cuffs off patient.
No patient who
truly required
intervention was
not correctly
identified
64
1. Patients receiving parenteral opioids should be monitored continuously
regardless of location in the hospital.
2. Minimal set of continuous data should include pulse, oxygen saturation,
respirations and expiratory carbon dioxide.
3. Monitoring should be a team effort involving floor nursing, respiratory
therapy and physicians, and all should be trained in the workflows and
technologies associated with continuous ventilation monitoring.
4. Notifications for clinical alarms related to hypoventilation concurrent with
hypoxia should be directed to nursing AND respiratory therapy.
5. Clinical or biomedical engineering should be part of the notification chain
for machine-specific alarms and errors that indicate possibility of false
readings.
6. When continuous surveillance is used, smart alarms using multiple
parameters that identify hypoventilation together with declining
oxygenation provide the best indicator of impending respiratory distress.
“The 6 Rights of OIRD Surveillance”
65
A Business & Clinical Case
for Continuous Surveillance
Thank you!
jzaleski@bernoullihealth.com